hcimonitord using more memory (5.3 rev3 on AIX 5.1)

[Rob Lindsey]

We just recently upgraded our CL system from 3.8.1P rev6 to 5.3 rev3.  One thing that I have noticed is that the monitord is using up a lot of memory over a short period of time.  When I checked the system this morning I found that all of my hcimonitord programs are using up 78K a piece.  I did an experiment of shutting down the monitord for 1 site and it went down to 5K.  The last time I stopped and started each monitord program for each site was Friday morning (3 days ago).  Has anyone else seen this type of memory usage in their monitord programs?

Rob

Topic

[Jay Yeiser]

I haven’t seen this.  Does your 5.3 version have a bunch of alerts in the Alert Monitor running that weren’t running on your old version?

[Rob Lindsey]

Not a buch.  Just the “default” ones plus ones that check the processes for being down.

[Anonymous]

Yes, we noticed that too with processes. Also, performance was degraded with time. In some cases, we had process panics after a high number of messages were processed. We never figured out why (everything is OK if you re-start the engine) …so we have scripts that will bounce the processes and monitord once a day when the queues are empty

[Kathy Young]

We’re about to make this very same upgrade…Now I’m worried. What’s the word from QuoVadx?

[Rob Abbott]

We are not aware of any memory leaks in monitord.  We ran several test cases with sites running for weeks with default alerts.  

I would appreciate it if anyone upgrading to rev3 would keep an eye on this and let us know if you see monitord growing without bound.  Please include detailed o/s information when you post results.

The last leak I am aware of was in the engine, fixed in (I believe) 5.3 rev1 and 5.2 rev2.  This leak would occur if you had cross-process routing going on, the message metadata would leak.  Normally not a lot of memory so this would take a while to be noticed.

Rob Abbott
Cloverleaf Emeritus

[Anonymous]

Rob and all,

I apologize; I should read the postings more carefully. We were in 5.1 when we noticed that issue and now that we are in 5.2 rev 2 the monitord leaks are gone,

[Terry Kellum]

I think that there’s some log file limiting functionality in there.  

Go to netconfig and pick the process menu, Configure.

Click “Automatic Log Cycling” and set a threshold in K.

If you are not overly attached to your logs, this works well.

[Author]

Replies